Nuclear Magnetic Resonance (NMR) is a powerful technique used to characterize fluids and flow in porous media. The NMR relaxation curves are closely related to pore geometry, and the inversion of the NMR relaxometry data is known to give useful information with regards to pore size distribution (PSD) through the relative amplitudes of the fluids stored in the small and large pores. While this information is crucial, the main challenge for the successful use of the NMR measurements is the proper interpretation of the measured signals. Natural porous media patterns consist of complex pore structures with many interconnected or “coupled” regions, as well as isolated pores. This connectivity along the throats changes the relaxation distribution and in order to properly interpret this data, a thorough understanding of the effects of pore connectivity on the NMR relaxation distribution is warranted. In this paper we address two main points. The first pertains to the fact that there is a discrepancy between the relaxation distribution obtained from experiments, and the ones obtained from solving the mathematical models of diffusion process in the digitized images of the pore space. There are several reasons that may attribute to this such as the lack of a proper incorporation of surface roughness into the model. However, here we are more interested in the effects of pore connectivity and to understand why the typical NMR relaxation distribution obtained from experiments are wider, while the numerical simulations predict that a wider NMR relaxation distribution may indicate poor connectivity. Secondly, by not taking into account the pore coupling effects, from our experience in interpreting the data, we tend to underestimate the pore volume of small pores and overestimate the amplitudes in the large pores. The role of pore coupling becomes even more prominent in rocks with small pore sizes such as for example in shales, clay in sandstones, and in the microstructures of carbonates.